Heterogeneous LTE Networks
and Inter-Cell Interference
Coordination
Volker Pauli, Juan Diego Naranjo, Eiko Seidel
Nomor Research GmbH, Munich, Germany
December, 2010
Summary
Initial deployments of LTE networks are based
on so-called homogeneous networks consisting
of base stations providing basic coverage, called
macro base stations. The concept of
heterogeneous networks has recently attracted
considerable attention to optimize performance
particularly for unequal user or traffic
distribution. Here, the layer of planned highpower macro eNBs is overlaid with layers of
lower-power pico or femto eNBs that are
deployed in a less well planed or even entirely
uncoordinated manner. Such deployments can
achieve significantly improved overall capacity
and cell-edge performance and are often seen as
the second phase in LTE network deployment.
This
paper
discusses
the
concept
of
heterogeneous networks as compared to
homogeneous networks. It demonstrates the
need for inter-cell interference coordination
(ICIC) and outlines some ICIC methods that are
feasible with release 8 /9 of the LTE standard.
System-level simulation results illustrate the
benefits of the various features discussed in the
following.
Introduction
As explained a macro-cell deployment is carefully
planned prior to its roll-out and some
optimization can be performed after roll-out to
get the best of such a system. Nevertheless,
future increase of performance in such
homogenous networks is limited particularly in
case of unequal traffic or user distribution.
An elementary and well-known strategy to
increase the capacity of a cellular network is to
reduce the cell size. The underlying effect is to
further increase the frequency reuse, also known
as “cell-splitting gain”. For cost optimization
different types of eNBs are used for different
purposes, e.g. large-scale eNBs for basic
coverage, smaller eNBs to fill coverage holes or
to improve capacity in hot-zones or at the
boundaries between large-scale eNBs’ coverage
areas, and possibly even smaller eNBs for indoor
coverage.
LTE is designed for a frequency reuse of 1,
meaning that every base station uses the whole
system bandwidth for transmission and there is
no frequency planning among cells to cope with
interference from neighbouring cells. Hence, LTE
macro-cell deployments experience heavy
interference at the boundaries of the cells.
Placing a new eNB between macro-cells would
boost the SINR levels for users located there,
achieving a more uniform user satisfaction and
overcoming link-budget constraints.
Nomor Research GmbH / info@nomor.de / www.nomor.de / T +49 89 9789 8000
1/9
Similarly, the deployment of eNBs inside
buildings is a reasonable strategy, as it fills
coverage holes that typically occur due to the
penetration loss imposed by the walls, while it
causes relatively little interference to the macro
network for the same reason. Hence, femto-cell
deployments are being investigated vigorously in
industry and in standardization bodies.
Since the backhaul connection of many small
size base stations at remote locations can
increase operational expenditures significantly,
alternative backhaul techniques using the LTE air
interface are a hot topic as well. The backhaul
connection to the base station, serving as relay
node in this case, can be on the same frequency
(in-band) or on different frequencies (out-band).
In-band relays have been standardised in LTE
Release 10 specification while the out-band
relays can be used without Release 8/9
specification. Another backhaul alternative is the
use of remote radio heads (RRH).
Layers of Heterogeneous Networks in LTE
In the previously described cases new types of
nodes need to be installed in addition to the
macro base stations resulting in a so-called
“heterogeneous network” (HetNet). The new
network elements might be pico- or femto- base
stations, Remote Radio Heads (RHH) and/or
relays nodes.
Figure 1: Heterogeneous multilayer cellular
environment
In Figure 1, such a multi-layer deployment is
illustrated, where a macro-cell deployment is
supported by several low-power nodes, all of
which solve specific issues as discussed above.
Classification into different types of nodes is not
always easy, although some function might be
specific to different types, e.g. home eNB
specific functions summarized in [1]. Base
stations of different output power are defined in
TS36.104 [5] with the following classification:
BS class
Wide Area BS
Local Area BS
Output power per Tx
antenna
no upper limit
< + 24 dBm (1 antenna)
< + 21 dBm (2 antennas)
< + 18 dBm (4 antennas)
Home BS
< + 20 dBm (1 antenna)
< + 17 dBm (2 antennas)
< + 14dBm (4 antennas)
The wide area or macro base station does not
have a limited output power. The local area base
stations or pico base stations on the other hand
are fully-featured eNBs with lower transmission
power, reduced size and cost that can be
deployed easily for improving conditions in
coverage holes and providing higher data-rates
at cell edge or in hot-spots. The home base
station or femto-cell has a limited power that is
actually lower than a terminal output power. The
restriction is applied to minimize the impact in
severe
interference
conditions
from
uncoordinated deployment. It is typically
operated in a closed mode, meaning that only
certain users that are part of a “closed
subscriber group” (CSG) are allowed to connect
to such an eNB.
Interference Scenarios in HetNets
Typically, pico-eNBs have a relatively small
coverage area due to their low transmit power.
If not placed specifically in a hot spot, only a
small number of UEs will connect to the pico-eNB
which will limit the gain from offloading the
traffic from the macro cells. Signalling that
supports load balancing between the macro and
the pico-eNB has been standardized and is
commonly used. This can be achieved by biasing
handover decisions between the different eNBs
Nomor Research GmbH / info@nomor.de / www.nomor.de / T +49 89 9789 8000
2/9
such that UEs are handed over to pico-eNBs
earlier than usual, thus shifting load from the
macro-eNB to the pico-eNB. As this corresponds
to an expansion of the range of the pico-eNB,
this feature is typically referred to as “range
expansion”. Similarly cell selection parameters
can be adapted for users that are idle.
Figure 2 shows an overview of a number of
frequency partitioning methods. These ICIC
methods describe basic rules on how a system
performance boost can be achieved by managing
the system bandwidth and transmit power. The
following discussion will introduce general
notions about frequency partitioning and the
available options for LTE release 8/9.
Range expansion is not only effective for
optimizing the use of resources in the system,
but also for reducing the frequency of handovers, hence improving system throughput and
user experience [2].
However, with range expansion the UEs handed
over to the pico-eNB suffer from lower than
usual SINRs, a situation that other techniques
like beamforming and power boosting would not
be able to completely resolve [2]. So, despite the
above benefits this feature presents a challenge
to interference management for HetNets, and
the deployment of appropriate ICIC algorithms
can further boost system performance.
Furthermore, when considering CSGs, yet
another challenge for interference management
arises if UEs are in the coverage area of a homeeNB, typically well shielded from the macro-eNB,
but are not allowed access to it. This creates
complex high-interference scenarios in both
transmission directions that cannot easily be
solved. In the downlink such a “macro-UE” is the
victim being exposed to heavy interference from
the home-eNB, whereas in the uplink the macroUE is the aggressor severely disturbing
transmissions to the home-eNB, cf. e.g. [2].
ICIC Methods
As motivated in the previous section, Inter-Cell
Interference Coordination (ICIC) plays a vital
role in heterogeneous networks. ICIC techniques
in LTE are mostly limited to the frequency
domain, e.g. only partial use of resources in
frequency direction and/or adaptation of power
levels.
Figure 2: Different inter-cell interference
coordination schemes
Full Frequency Reuse
Full Frequency Reuse means that no frequency
partitioning is performed between eNBs of the
same network. eNBs in this configuration
transmit with uniform power over the entire
system bandwidth. This is the conventional way
of operating an LTE network. The main pitfall to
this configuration is that cell-edge users
experience heavy interference from neighbouring
cells in the downlink and create heavy
interference in the uplink, greatly degrading their
communication performance.
Hard Frequency Reuse
This ICIC method is typically seen in GSM
networks, when it comes to distribution of
frequencies among the cells. When applied to
LTE it means that the sub-carriers are divided
into 3, 4 or 7 disjoint sets. These sets of subcarriers are assigned to the individual eNBs in
such a way that neighbouring cells don't use the
same set of frequencies. This reduces the
interference at the cell edge of any pair of cells
Nomor Research GmbH / info@nomor.de / www.nomor.de / T +49 89 9789 8000
3/9
significantly and can be considered the opposite
extreme to Full Frequency Reuse in matters of
frequency partitioning techniques. While user
interference at cell edge is maximally reduced,
the spectrum efficiency drops by a factor equal
to the reuse factor.
SINR levels for cell-edge UEs in the hightransmit power regions, and the other regions
still have suitable SINR levels for cell-center UEs,
as it is depicted in Figure 3. It can nicely be seen
that the high power region serves a larger
coverage area. This scheme is particularly useful
for ICIC in the downlink.
Fractional Frequency Reuse
This is a hybrid frequency partitioning combining
the concepts of the two previous schemes. It
consists of dividing the spectrum into two parts
which will have different frequency reuse. One
section of the system spectrum is used in all
cells, while the other part of the spectrum is
divided among different eNBs as in hard
frequency reuse. The idea is that the eNB would
assign the fully-reused frequency chunks to
center-cell UEs and the other chunks to the celledge UEs. This scheme is particularly useful for
ICIC in the uplink, where severe interference
situations can occur where the user is located
close to a strong interferer in the neighbor cell.
Soft Frequency Reuse
In this frequency partitioning method an eNB
transmits in the whole system bandwidth, but
using a non-uniform power spectrum.
Figure 2 (bottom) illustrates power spectrum
assignments in the different cells of a system
with Soft Frequency Reuse and reuse 3. It can
be noticed that in the spectrum there is a region
of high-power transmissions and some regions of
low-power transmissions. Using a similar
strategy as the previous method, resources in
the high-power region are preferably assigned to
UEs located at the cell edge, while cell-center
UEs are typically assigned resources in the lowpower regions.1
This coordination scheme leads to improved
1
Note, however that UEs can only be
assigned to either of the regions and
reassignments require RRC signaling.
Figure 3: SINR strength distribution over the cell
with soft frequency partitioning.
Note that in the case of HetNets it may also be
useful to block some of the resources available in
frequency direction entirely at the pico or femto
eNBs but let the macro eNB use the entire
spectrum. It grants resources to the macro cells
that are not interfered by the smaller eNBs, but
comes at the price of reduced throughput for the
small eNBs. It does not necessarily result in a
reduced user throughput since the number of
UEs connected to the small eNBs is usually
limited.
X2 Signalling to support ICIC
The already discussed partitioning schemes can
in principle be easily implemented in the radio
resource management (RRM) of the eNB. They
can be configured statically and run without
interaction between different eNBs. However,
better performance can be achieved if the above
schemes are configured dynamically, based on
Nomor Research GmbH / info@nomor.de / www.nomor.de / T +49 89 9789 8000
4/9
active information exchange between eNBs in
order to better adapt to the current state of the
network. X2 signalling provides the mechanisms
for the exchange of this information, cf. [3] and
references therein.
Note that the signalling and its information
elements are well defined in 3GPP specifications,
but the specific reaction of an eNB in response to
the reception of such a message is not. Hence,
in a multi-vendor network, there is no
guaranteed reaction of an eNB to an incoming
ICIC-related message from another eNB.
The following sections describe a set of X2
messages meant for ICIC in Release 8/9.
Relative Narrowband Transmit Power Indicator,
RNTP
This information message is sent to neighbour
eNBs. It contains 1 bit per physical resource
block (PRB) in the downlink, indicating if the
transmission power on that PRB will be greater
than a given threshold. Thus, neighbour eNBs
can anticipate which bands would suffer more
severe interference and take the right scheduling
decisions immediately rather than relying on the
UEs' CQI reports only.
sent out proactively by eNBs, this indicator is
only triggered when high-interference in the
uplink direction is detected by an eNB.
The
overload indication will be sent to neighbour
eNBs whose UEs are potentially the source of
this high interference. The message contains a
low, medium or high interference level indication
per PRB. The question, which cell is responsible
for the high interference is of course not a trivial
one.
It can be seen that specifications for Release 8/9
were defined having mostly homogeneous
macro-cell deployments in mind. It provides only
relatively simple means for ICIC and fails for
instance to address mechanisms to improve
inter-cell interference on control channels, e.g.
PDCCH, PHICH and PCFICH. This limits for
example the aggressiveness in which range
expansion can be configured and hence the
gains that can be achieved using HetNets with
range expansion. Release 10 will enhance the X2
signalling in order to support more sophisticated
ICIC
algorithms,
involving
interference
coordination among eNBs in time domain [6].
Nomor Research will report in one of its next
newsletter
about
progress
in
3GPP
standardisation.
High Interference Indicator, HII
This indicator for uplink transmissions works
similarly to the previous RNTP message for the
downlink. There is one bit per PRB indicating if
neighbouring
eNBs
should
expect
high
interference power in the near future. Hence,
typically only PRBs assigned to cell-edge UEs are
indicated. RSRP measurements as part of
handover measurement reports can identify celledge UEs. In a similar way this indicator can be
used to identify the bands used in a frequency
partitioning scheme.
Simulation Results
In this section, we present some system-level
simulation results to illustrate the effects of the
various features described above. For this
purpose 3 scenarios are evaluated. The first
scenario
corresponds
to
a
macro-cell
deployment. The second scenario adds a layer of
pico-eNBs on top of the macro-cell deployment.
The third scenario includes Range Expansion,
additionally to the HetNet layout. Range
expansion is achieved by setting the handover
biases between pico-eNBs and macro-eNBs to 4
dB.
Interference Overload Indicator, OI
While the previously described X2 messages are
These scenarios are configured according to the
guidelines for Case 1 described in the 3GPP
Nomor Research GmbH / info@nomor.de / www.nomor.de / T +49 89 9789 8000
5/9
technical reports for simulation of HetNets [4].
In summary, the simulations are fully dynamic
with full RRM functionality such as scheduling
(simple proportional fair strategy in this case),
link adaptation, hybrid ARQ, etc. Furthermore,
the users are moving with 3km/h, and they carry
out handovers at the cell boundaries as specified
in 3GPP (handover margin/hysteresis is set to
3dB). 30 UEs are dropped per macro-cell area
according to configuration 4a, where 2 UE
clusters consisting of 2 UEs each are located in
each macro-cell and a pico-eNB is placed in the
middle of every UE cluster. The remaining UEs
are dropped uniformly in the macro-cell area.
Figure 4 illustrates an example of such a
deployment.
second improvement in throughput is achieved
when range expansion (RE) is used. By the
handover of macro-UEs to pico-eNBs traffic is
offloaded from the macro-cell layer to a number
of pico-eNBs.
From Figure 6 it can be concluded that the
introduction of RE in the system lowers the
effective SINRs.
Figure 5: PDCP Throughput, Configuration 4a with
2 UE clusters per macro
Figure 4: Cell layout for configuration 4a and 2
pico-eNBs per macro-cell scenario.
Figure 5 and 6 show the CDFs of PDCP
throughput and Effective SINR results obtained
in this scenario, respectively.
The first observation is the improvement of
throughput achieved by the introduction of a
pico-cell layer to the macro-layer scenario. A
Figure 6: Effective SINR CDF, configuration 4a with
2 UE clusters per macro.
This happens because of the high interference
experienced by UEs in the extended pico-cell
area. This is however more than compensated
by the fact that the UEs connected to the picoeNBs get significantly more resource than when
Nomor Research GmbH / info@nomor.de / www.nomor.de / T +49 89 9789 8000
6/9
connected to the macro-eNB due to a lower load.
Furthermore, we evaluated a simple static ICIC
scheme for downlink. Here, frequencypartitioning is performed in a way that macroeNBs transmit in the full frequency spectrum (50
PRBs) while pico-eNBs transmit in a reduced
band (30 PRBs), which means that the pico-eNB
is not generating interference in the remaining
unused downlink resources. Hence a frequency
selective scheduler at the macro-eNB, after
evaluating the measurement reports, would take
advantage of this effect. The macro-eNB would
assign the low-interference frequency segments
to macro-UEs located close to the pico-cells,
protecting
these
macro-UEs
from
the
interference from the respective pico-eNB. In
this scenario it can be observed that no
significant variation is experienced at the PDCP
throughput CDF, although a slight increase in the
SINR is seen, indicating a trade off between
resources and transmission power that renders
in same throughput performance for pico-cell
UEs. Figure 7 shows the TPs for the RE scenario
with and without this ICIC strategy.
ICIC technique is that the optimal configuration
depends very much on the specific scenario, i.e.
the location of the pico-eNB in the macro-cell
and on the distribution of UEs and the traffic in
the vicinity of the pico-eNB. A semi-static
approach continuously adapting the frequency
partitioning on a slow basis provides further
potential for performance enhancements.
Configuration 4b [4] with 2 pico-eNBs was also
simulated under the same RE+ICIC scenario.
The basic difference between configuration 4a
and 4b is the dropping of UEs. In Configuration
4b 10 UEs are dropped per UE cluster, but the
total number of UEs is still set to 30 per macrocell. Figure 8 shows an example of such a
scenario, where the higher density of UEs
around pico-eNBs is visible compared to
configuration 4a in Figure 4.
Figure 8: Cell layout for configuration 4b and 2
pico-eNBs per macro-cell scenario.
Figure 7: PDCP Comparing effects on TP due to
ICIC strategy, configuration 4a.
It can be concluded that severe interference
impact from pico- or femto-cells onto the macrocells in HetNet deployments could be resolved
using this technique. A problem with this static
Figure 9 shows the results for this configuration.
In this case, the number of UEs connected to the
pico-eNBs and hence the offloading gain from
introducing pico-eNBs is already very significant.
Expanding the range of the pico-eNBs actually
degrades the performance as the pico-eNBs
Nomor Research GmbH / info@nomor.de / www.nomor.de / T +49 89 9789 8000
7/9
become overloaded. Similarly, using statically
configured ICIC as described above will further
reduce the resources available for the pico-UEs
leading to overall performance degradation.
Here, a dual ICIC scheme, leaving some
resources to the pico-UEs only would be more
beneficial.
in an upcoming
Research.
newsletter
from
Nomor
References
[1] Nomor Newsletter, “LTE Home Node Bs and
its enhancements in Release 9”.
[2] R1-100700, Interference conditions in
Heterogeneous Networks. 3GPP TSG RAN 1
meeting #59 bis.
[3] TS 36.420, E-UTRAN; X2 General Aspects
and Principles; Release 9, December 2009.
[4] TS 36.814, E-UTRAN; Further advancements
for E-UTRA physical layer aspects; Release 9,
March 2010.
[5] TS 36.104 E-UTRA; Base Station (BS) radio
transmission and reception; Release 9, Oct 2010
Figure 9: Comparing effects on TP due to ICIC
strategy, configuration 4b.
So again, it becomes clear that both range
expansion and ICIC must be configured
dynamically depending on the scenario to
achieve best performance.
Conclusions
Heterogeneous Networks constitute a promising
concept to improve system performance, to
mitigate coverage holes, to provide a uniform
user experience over the entire cell area and last
but not least to satisfy high traffic demands in
hot-zones. Properly configured range expansion
has the potential to increase the gains as it leads
to better load balancing over the different
network layers. The applicability of range
expansion is, however, limited by inter-cell
interference calling for advanced ICIC methods.
Unfortunately, support of ICIC is limited to data
channels in LTE release 8 / 9. Further
enhancements of ICIC support particularly for
control channels are in the process of
standardisation and will be included in Release
10 specification. A related report will be provided
[6] RP-101229 RAN2 CRs on Core part:
Enhanced ICIC for non-CA based deployments of
heterogeneous networks for LTE RAN2;
3GPP
TSG RAN meeting #50
LTE Discussion Forum
http://www.linkedin.com/groups?gid=1180727
Note: This white paper is provided to you by Nomor
Research GmbH. Similar documents can be obtained from
www.nomor.de. Feel free to forward this issue in
electronic format. Please contact us in case you are
interested in collaboration on related subjects.
Disclaimer: This information, partly obtained from
official 3GPP meeting reports, is assumed to be
reliable, but does not necessarily reflect the view of
Nomor Research GmbH. The report is provided for
informational purpose only. We do not accept any
responsibility for the content of this newsletter. Nomor
Research GmbH has no obligation to update, modify or
amend or to otherwise notify the reader thereof in the
event that any matter stated herein, or any opinion,
projection, forecast or estimate set forth herein,
changes or subsequently becomes inaccurate.
Nomor Research GmbH / info@nomor.de / www.nomor.de / T +49 89 9789 8000
8/9
System Level Simulation Services
Nomor Research has developed a comprehensive
simulation environment with fully compliance to
standards such as LTE, LTE Advanced and HSPA
and offers related services to support research,
development and standardisation.
Features of the dynamic multi-cell, multi-user
system level simulator include:
• macro-cell and HetNet deployments (pico-,
femto-cell, relay nodes)
• flexible base station and user configurations
and drop models
• different transmitter and receiver chains incl.
MIMO, ZF, MMSE
• channel modeling with slow/fast fading,
pathloss, full user mobility
• intra- and intercell interference modeling for
OFDMA, SC-FDMA and WCDMA
• simultaneous up- and downlink simulation
• 2D and 3D antenna pattern & beam forming
• Extensive metrics and KPIs: capacity,
throughput, spectral efficiency, user QoS etc
• High-speed processing close to real-time
using multi-core & graphic card processors
The simulator can be used on project basis or in
customized
simulation
campaigns.
The
performance of the system level simulator has
been calibrated to simulation results obtained in
standardisation.
Research on advanced algorithms include, but
are not limited to:
• advanced features as link adaptation, HARQ,
power control, measurements
• scheduling
and
resource
allocation
algorithms considering channel and buffer
status, QoS etc.
• inter-cell
interference
coordination,
avoidance and cancellation
• single user-, multi-user MIMO with open and
closed loop feedback
• cooperative multi-point transmission and
reception
• functions for self-organising and selfoptimizing networks (e.g. load balancing,
mobility optimization, tilt optimisation, range
extension, power saving etc. )
• radio resource management for heterogeneous networks
If you are interested in our services please
contact us at info@nomor.de or visit us at
http://www.nomor-research.com/simulation
Nomor Research GmbH / info@nomor.de / www.nomor.de / T +49 89 9789 8000
9/9